Geological framework for natural hydrogen generation in the Jeju basin and Hupijiao Rise, Northern East China Sea

1 Introduction

The transition to low-carbon energy systems has increased interest in alternative hydrogen resources. Natural hydrogen (also referred to as “white” or geological hydrogen) is recognized as a potential resource that substitutes conventional hydrocarbon energy (Lévy et al., 2023; Zhao et al., 2023; Jackson et al., 2024). Natural hydrogen can be generated through a variety of abiotic and biotic processes, including serpentinization, water–rock interactions (WRI), radiolysis of water, mechanochemical reactions associated with fault and rock fractures, thermal decomposition of overmature organic matter, microbial metabolism, and mantle degassing. The relative contributions of these processes vary depending on lithology, temperature, fluid availability, redox conditions, tectonic activity, and subsurface residence time.

Natural hydrogen exploration has expanded rapidly, although sustained production remains limited. The Bourakebougou field in Mali is known for hydrogen-rich gas that has been continuously used for local electricity generation without an apparent decline in reservoir pressure (Prinzhofer et al., 2018; Maiga et al., 2023). The Ramsay Project in South Australia has identified hydrogen and helium bearing intervals, followed by appraisal drilling and well-testing programs (Boreham et al., 2021; Gold Hydrogen Ltd., 2023; Gold Hydrogen Ltd., 2026). In the Lorraine Basin of France, dissolved H₂ concentrations increase with depth, which has led to deeper appraisal drilling (Pironon and de Donato, 2023; Rashid et al., 2025). Hydrogen occurrences have also been revealed in the Bulqizë ophiolite of Albania and the Precambrian basement of the Nemaha region, USA (Guélard et al., 2017; Touré et al., 2026). However, early production testing at Nemaha has not demonstrated sustained hydrogen flow (HyTerra Ltd., 2026). These cases show that a hydrogen system requires a source, connected migration pathways, reservoirs, seals, and preservation from leakage or consumption.

The study area covers the Jeju Basin and Hupijiao Rise in the northern East China Sea Shelf Basin (ECSSB). The basin comprises Mesozoic–Cenozoic sub-basins separated by uplifted zones of Precambrian crystalline basement (Zhou et al., 1989). Recent geodynamic modeling identifies rift-inversion settings as potential hot spots for natural H₂ generation, because inversion reheats rift-stage source rocks into the optimal H₂ generation window and reactivates basement-involved faults as migration pathways (Zwaan et al., 2025). These modeling results are comparable to the multistage rifting and inversion history of the Jeju Basin (Lee et al., 2006; Li et al., 2009). The granitic basement may support radiolytic H₂ generation, while mafic intrusions may provide ferrous iron for water–rock interactions. Deep faults may link these source domains to reservoirs and seals in the sedimentary deposits.

The dataset for the study area comprises 2D and 3D seismic reflection data, wireline logs, and geochemical data derived from analyses of drill cuttings, all of which were acquired during hydrocarbon exploration. Their reinterpretation for hydrogen exploration requires emphasis on source rocks and migration pathways rather than on conventional reservoir rocks (Zhang and Li, 2024). Magnetotelluric modeling in the southern Gawler Craton provides a recent example. Crustal resistivity and major faults were used to evaluate possible generation domains and migration pathways (Margiono et al., 2026). The available datasets can be used to identify fractured and water-bearing basement, radiogenic granites, Fe-bearing intrusions, fault networks, and reservoir-seal relationships.

The primary objective of this study is to establish a first-order, conceptual evaluation framework for subsurface natural hydrogen systems in frontier basins, using the available geophysical and geochemical data from the Jeju Basin and Hupijiao Rise. As the accessibility of aqueous fluids is a prerequisite for both radiolysis and serpentinization, the characterization of fault and fracture systems is critical. This study applied artificial intelligence (AI) based fault interpretation using convolutional neural networks (CNNs) to seismic attribute volumes, which preserves structural continuity in complex, intrusion affected basements more reliably than filter-based attributes (Han and Cader, 2020; Zhang and Li, 2024). Wireline logs are used to identify fractured, water saturated basement intervals, and Rock-Eval pyrolysis data are evaluated against recent kinetic models to assess H₂ potential from the thermal decomposition of organic matter. This multidisciplinary analysis provides a structured, conceptual foundation for future quantitative resource assessments in continental margin basins.

2 Geological setting

The East China Sea Shelf Basin (ECSSB) is a Mesozoic–Cenozoic superimposed basin developed on a basement composed primarily of Proterozoic–Palaeozoic metamorphic and igneous rocks (Yang et al., 2020). The crystalline basement is composed of gneiss, schist, and Cretaceous granite intrusions, which were subsequently deformed during Mesozoic–Cenozoic tectonic processes. These rocks form the acoustic basement beneath several sub-basins, including the Jeju Basin and Hupijiao Rise (Yang et al., 2004; Li et al., 2009) (Figure 1).

In the Jeju Basin, the acoustic basement is unconformably overlain by fluvial to lacustrine sedimentary successions composed of sandstone, shale, and volcaniclastic deposits. These sediments were primarily sourced from adjacent continental and uplifted regions. The Hupijiao Rise, a structural high on the western flank of the Xihu Depression, is composed of granite, gneiss, Proterozoic schist, and mylonite, which are intruded by widespread Miocene basaltic dikes and sills, reflecting intense post-rift magmatism (Han et al., 2015; Song et al., 2024).

The ECSSB formed through interactions between the Eurasian, Pacific, and Philippine Sea plates. Seismic interpretation, geophysical modeling, and stratigraphic data indicate that the basin evolved from a Triassic passive margin, into a Jurassic compressional foreland, followed by Cretaceous–Paleogene extensional rifting and back-arc basin formation (Li et al., 2009; Yang et al., 2020; Weilin et al., 2019). This tectonic conversion from a Tethyan to Pacific regime restructured the basin architecture, giving rise to deep, steeply rooted fault systems that facilitate fluid migration and crustal thinning that can increase heat flow and potential hydrogen generation (Donzé et al., 2020; Etiope, 2023).

The Xihu Depression, the largest hydrocarbon bearing area within the ECSSB, lies in the central-eastern part of the basin (Figure 1). The western slope of the depression is bounded by the Hupijiao Rise, which served as a persistent sediment source from the Eocene to Miocene and preserves geochemical signatures of basement derived provenance (Song et al., 2024).

According to previous studies, the ECSSB basement is an extension of the South China Block (Yang et al., 2020). It is expected to host granitic intrusions characterized by high abundances of U, Th, and K, which are essential for radiolytic hydrogen production. The presence of mafic-ultramafic basement rocks, extensive igneous intrusions, active faulting, and overlying sedimentary deposits provides essential elements for the generation, migration, and accumulation of hydrogen.

3 Data and methods

The dataset used in this study comprises 2D and 3D seismic reflection data, wireline logs, and geochemical data from drill cuttings. The seismic data were acquired by the Korea Institute of Geoscience and Mineral Resources in the Jeju Basin and the Hupijiao Rise. The 2D seismic dataset provides regional coverage of the study area, whereas the 3D seismic volume covers 79 km² of the Jeju Basin. The 3D volume has a bin spacing of 25 × 12.5 m and a dominant frequency of approximately 52 Hz at the basement level. Wireline logs and drill cuttings were obtained from two exploration wells, Well A and Well B, located in the Jeju Basin (Figure 2A). Geochemical data for the drill cuttings were compiled from Kim and Son (2013).

Seismic interpretation was conducted using Petrel (SLB, 2025) to delineate basement structure, and igneous intrusions such as sills and dikes. The 3D seismic volume was used to interpret basement geometry, faults, and igneous intrusions, and to perform seismic-attribute analysis. Basement-involved faults were detected using the AI-based fault-detection workflow in Geoteric. The workflow employs a pretrained multilayer convolutional neural network (CNN) for fault delineation (Han and Cader, 2020). The 3D seismic volume was preconditioned using structure-oriented filtering to suppress random noise while preserving structural discontinuities. The pretrained network was then applied without additional fine-tuning to generate a fault-detect volume. Discrete fault surfaces were extracted using a probability threshold of 0.5 and a minimum patch size of 3. Dip and azimuth were calculated from the extracted surfaces, and rose diagrams were constructed separately for the sedimentary basin and volcanic structural high. The extracted faults were compared with manually interpreted faults on selected 2D seismic lines to assess the possible misclassification of intrusion margins and stratigraphic terminations. The fault-detection volume was also co-rendered with a CMY (cyan, magenta, and yellow) blended frequency decomposition volume to examine fault trends and igneous features.

The wireline dataset includes gamma-ray, density, resistivity, and neutron porosity logs from Wells A and B. These logs were interpreted to constrain lithology following Rider and Kennedy (2011). The log responses were also used to identify fractured and water saturated intervals. These intervals indicate pore-water availability for radiolysis and water–rock interaction (Lin et al., 2005; Dzaugis et al., 2016).

The drill-cutting dataset comprises 32 samples from Well A and 33 samples from Well B, representing shale and coal intervals. According to Kim and Son (2013), the samples were analyzed using a Rock-Eval Turbo six instrument to determine kerogen type and thermal maturity. Elemental analyses using a LECO CHN-900 analyzer were used to quantify bulk nitrogen and the elemental composition of kerogen. The Rock-Eval and elemental data were used to evaluate the potential for H₂ generation from organic matter.

4 Results

Interpretation of seismic reflection data from the study area reveals that the basement structure can be subdivided into three parts: sedimentary basin (Jeju Basin), igneous intrusion, and the uplifted basement high (Hupijiao Rise) (Figures 2B-D). The Hupijiao Rise forms a regional structural high that separates the Jeju Basin from the Ieodo (Socotra) Basin in the northern ECSSB. The Hupijiao Rise has an average water depth of ∼40 m and an interpreted acoustic basement at approximately 800–1200 m.

The Jeju Basin is characterized by graben and half-graben structures, filled with fluvial and lacustrine sedimentary successions as well as volcaniclastic deposits (Figure 3). The thickness of the sedimentary deposits ranges from 400 to 3650 m, showing a significant contrast between uplifted zones and sedimentary depocenters. The high amplitude reflector represents an irregular unconformity surface that is locally disrupted by igneous intrusions (Figure 3). Seismic attribute analysis combined with CNN-based fault interpretation was applied to the 3D seismic volume to delineate seismic-scale faults and igneous intrusions. Fault extraction indicates that NNE–SSW-striking faults dominate the study area (Figures 4A,C). In the sedimentary basin, dip azimuths cluster between 11.25° and 45° (Figure 4D), whereas the volcanic structural high shows a broader distribution (Figures 4B,E). Integration of the fault-detection volume with the CMY frequency-decomposition attribute improves fault visualization and identification of igneous features. Despite the limited 3D seismic coverage, basement-involved fault patterns vary across fault zones and igneous complexes.

In Well A, cutting samples identified two granodiorite intervals and one granite interval. Granodiorite shows high gamma-ray, density, responses with low neutron values, whereas granite shows higher neutron and lower gamma-ray responses (Figure 5). The gamma-ray–neutron cross-plot shows low, scattered neutron values for granodiorite and higher neutron values with lower gamma ray for granite (Figure 5A).

The geochemical analyses were conducted on cutting samples from the sedimentary succession overlying the acoustic basement. Sandstone and volcaniclastic intervals contained little organic contents and were excluded from further organic geochemical interpretation. Elemental analysis of kerogen from cutting samples shows that the organic matter predominantly belongs to Type II and Type III kerogen, as illustrated in the Van Krevelen diagram (Figure 6A). Specifically, the hydrogen (H) content ranges from 1.49 to 5.93 wt% in Well A and from 2.04 to 5.89 wt% in Well B (Figure 6C). Downhole variations indicate that organic carbon (Org C), hydrogen, and oxygen contents fluctuate with depth across both wells (Figure 6B–D), reflecting changes in organic matter input and thermal maturity. Samples with low hydrogen content (1–3 wt%) likely indicate depleted or low hydrogen organic matter, while those with higher hydrogen content (5-6 wt%) indicate relatively hydrogen-rich kerogen. Rock-Eval results show TOC values of 0.12–7.13 wt% and T_max values of 397 °C–466 °C in Well A, compared with 0.13–4.18 wt% and 385 °C–439 °C, respectively, in Well B. In Well A, T_max exceeds 435 °C at ∼2,140 m and ∼2,400 m, and locally rises above 460 °C. In contrast, in Well B, T_max surpasses 435 °C at ∼2,150 m but remains below 440 °C to ∼2,700 m, indicating a slightly lower thermal stress compared to Well A (Figure 7). Hydrogen Index (HI) values range from ∼40 to ∼568 mg/g TOC in Well A and from <36 to ∼460 mg/g TOC in Well B, with the lowest values (<100 mg/g TOC) of Well A below 2,100 m (Kim and Son, 2013). These wells were drilled on a structural high, and their time-equivalent strata in the surrounding depocenters are likely buried below 2,800 m. At such depths, thermal maturity would exceed the hydrocarbon generation window (Hwang et al., 2026), with T_max values potentially entering the post-mature stage favorable for free H₂ generation.

5 Discussion

5.1 Radiolysis and water–rock interaction potential

The Hupijiao Rise is underlain by granitic–gneissic crystalline basement and overlain by Early to Middle Miocene sedimentary strata (Han et al., 2015). Previous studies indicate Miocene volcanism and persistent elevated thermal regime in the ECSSB, with regional heat flow reaching approximately 78 mW m^-2 since the Middle Eocene (Yao et al., 2026). The formation of the Hupijiao Rise during the Middle to Late Miocene is inferred to have enhanced basement fracturing and heat flow, increasing the potential for H₂ generation through radiolysis and water–rock interaction.

Volcanic sills and dikes are widely distributed in the northeastern ECSSB and are associated with asthenospheric upwelling and crustal thinning. These intrusions may enhance fracture permeability, fluid circulation, and mineral–fluid interaction, thereby supporting H₂ generation (Dzaugis et al., 2016). They may also contribute to trapping where intrusive bodies or altered zones reduce permeability, as reported from the Taoudeni Basin (Prinzhofer et al., 2018; Maiga et al., 2023). In volcanically influenced basins affected by mantle upwelling, part of the generated hydrogen may be converted to abiotic methane (Liu et al., 2025).

The global oceanic lithosphere is estimated to produce approximately 1 × 10¹¹ mol of molecular hydrogen per year through water-rock interaction (Bach and Edwards, 2003; Sleep and Bird, 2007; Kasting and Canfield, 2012). When both radiolysis and water-rock interactions are considered, H₂ production from the Precambrian continental crust is estimated at 0.73 to 4.60 million tons per year (0.36–2.27 × 10¹¹ mol yr^-1) (Lollar et al., 2014). These estimates highlight the potential significance of deep geological processes for natural hydrogen generation.

Artemieva et al. (2017) report typical ranges of 1.3–7.2 ppm for U, 5.6–28 ppm for Th, and 1.5–4.23 wt% for K in granitic-metamorphic basement. Using the radiolytic H₂ production model of Lin et al. (2005) and Dzaugis et al. (2016), and assuming average concentrations of 5 ppm U, 15 ppm Th, and 2.5 wt% K, we estimate a potential H₂ generation rate of approximately 2.77 µmol H₂ m^-3 yr^-1 for water-saturated crystalline basement. The radiolytic production rate is explicitly defined on a bulk rock volume basis (i.e., per cubic meter of bulk basement including both solid matrix and pore space). This rate aligns with globally reported values for continental crust, suggesting persistent low-flux production of molecular hydrogen through radiolysis (Lévy et al., 2023; Gluyas et al., 2024). Although radiolytic H₂ generation is energetically inefficient, sustained production over geologic timescales and across large rock volumes may nevertheless yield appreciable cumulative quantities of H₂.

In this study, the cumulative H₂ generation potential of the Hupijiao Rise is represented by radiolytic and WRI contributions. Mantle-derived H₂ is considered qualitatively but excluded from the mass balance because site-specific constraint on mantle H₂ flux is unavailable (Equation 1):

M H 2 gen = M H 2 , rad + M H 2 , WRI (1)

The radiolytic contribution is estimated from the rate of H₂ production through pore-water radiolysis induced by ionizing radiation emitted by uranium (U), thorium (Th), and potassium (K) (Equation 2):where R _rad is the radiolytic H₂ production rate (mol m⁻³ yr⁻¹), V_rock​ = Ah is the modeled basement volume (m³), A is area, h is basement thickness, Δt is the integration time (yr), and M _H₂​​​​ is the molar mass of H₂ (2.016 × 10⁻³ kg mol−1). The factor 10^–9 converts kilograms to megatonnes. The WRI contribution represents the stoichiometric H₂ generation capacity associated with oxidation of Fe(II)-bearing silicate minerals, expressed on a whole-rock FeO basis (Equation 3):where V_react ​ is the volume of reactive mafic rock, ρ is rock density, C_FeO​ is the mass fraction of reactive FeO, ηΔt​ is the fraction of reactive FeO oxidized over the integration interval, and K_conv is the stoichiometric conversion factor. For the idealized reaction 3FeO+H₂O→Fe₃O₄+H₂, K_conv = 9.35 × 10⁻³ kg H₂ kg⁻¹ FeO.

M H 2 , rad = R rad × V r o c k Δ t M H 2 × 10 − 9 (2)

M H 2 , WRI = V react ρ C FeO η Δ t K conv × 10 − 9 (3)

The Hupijiao Rise comprises metamorphic basement intruded by mafic igneous rocks, providing reactive ferrous iron favorable for low-temperature water-rock interaction (Song et al., 2024). As direct radioelement data for the Hupijiao Rise basement are unavailable, the radiolytic term was approximated using U, Th, and K contents (U = 2.4 ppm, Th = 9.12 ppm, K = 1.85 wt%) measured in Well A and B on an adjacent structural high.

Under the adopted parameter ranges, the modeled radiolytic contribution is approximately 6 Mt H₂, whereas the WRI contribution ranges from 7 to 125 Mt H₂. WRI becomes the dominant modeled source term as the reactive mafic-rock fraction and reaction extent increase (Figure 8). Mafic intrusions provide reactive Fe(II)-bearing substrates, whereas intrusion related fractures and contact zones may enhance water access and fluid migration. Crustal thinning and basement faulting could also facilitate the upward transfer of mantle-derived volatiles, including H₂, CO₂, and He, as documented in the Songliao Basin (Liu et al., 2025), although this contribution remains unquantified. Direct measurements of basement radioelement concentrations, Fe(II) speciation, intrusive-rock volume, alteration extent, effective fracture porosity, formation-water availability, and H₂ concentrations and fluxes are required to constrain the relative contributions of these processes.

5.2 Hydrogen generation potential from thermal decomposition of organic matter

The evaluation of hydrogen generation potential from thermal decomposition of organic matter requires a distinction between hydrogen bound in kerogen and free molecular hydrogen released during thermal maturation. Mahlstedt et al. (2022) reported that overmature organic matter generates free molecular hydrogen with cumulative potentials reaching 20 mg/g TOC, with peak generation at vitrinite reflectance (R_o) values of 3.5%–5.0%, marking the transition from the dry gas window to the post-mature stage.

To systematically link Rock-Eval parameters to H₂ release potential, we propose a four-stage classification scheme (Table 1) that integrates T_max, equivalent R_o (after Jarvie et al., 2005), Hydrogen Index (HI), and kerogen-bound hydrogen content. The boundaries are fixed to standard thermal maturity thresholds (Tissot and Welte, 1984; Peters and Cassa, 1994). The scheme distinguishes (i) a retention stage (T_max < 435 °C) where hydrogen is maximally retained in kerogen, (ii) a mixed generation stage (435 °C–470 °C) in which H₂ is co-generated with hydrocarbons but is largely consumed by hydrocarbon radicals through clay mineral catalysis, (iii) an active release stage (470 °C–590 °C) in which progressive aromatization and condensation begin to mobilize structurally bound hydrogen via amorphous-carbon catalysis, and (iv) a peak liberation stage (>590 °C) corresponding to the post-mature free H₂ window of Mahlstedt et al. (2022), in which molecular hydrogen is generated after CH₄ generation has finished. The R_o–T_max conversion follows Jarvie et al. (2005), calibrated for Type II kerogen; extrapolation beyond T_max ∼500 °C therefore carries increased uncertainty.

TABLE 1. Four-stage classification scheme linking Rock-Eval thermal maturity (T_max) to molecular-hydrogen (H₂) release potential in source rocks (Hanson and Hanson, 2024; Jarvie et al., 2005; Mahlstedt et al., 2022).

In the Jeju Basin, Rock-Eval data from Wells A and B record localized intervals of elevated thermal stress. Well A exhibits T_max values exceeding 460 °C at ∼2,140 m and ∼2,400 m, placing these intervals at the upper limit of the mixed-generation stage and approaching the active-release threshold (∼470 °C; equivalent R_o ≈ 1.3%). The accompanying decline in HI to <50 mg/g TOC and the reduced elemental hydrogen content (2.0–3.3 wt%; Figure 6C) are consistent with the geochemical signatures expected for this transitional stage, supporting the internal consistency of the classification. Well B remains predominantly within the mixed-generation stage (T_max 435 °C–440 °C), indicating spatially heterogeneous thermal overprinting. Both wells were drilled on a structural high, and their time-equivalent strata are below 2,800 m. At such depths, thermal maturity may exceed that recorded on the structural high, potentially reaching the active-release stage and locally entering, the post-mature window for free H₂ generation. The elevated T_max values in Well A reflect transient thermal overprints related to igneous intrusions and asthenospheric upwelling. This environment parallels the deep Songliao Basin, where localized magmatic heating promotes thermal decomposition of organic matter, facilitating H₂ release (Horsfield et al., 2022; Liu et al., 2025). However, the ECSSB has a different continental-margin setting and post-rift magmatic regime that may affect the timing and distribution of H₂ generation. The decrease in HI and increase in T_max indicate progressive carbonization. However, the maturity of these intervals (R_o ∼1.1%) remains below the post-mature free-H₂ window (R_o 3.5%–5.0%), indicating mobilization of structurally bound hydrogen rather than free-H₂ expulsion.

Some uncertainty remains because the geochemical data were obtained from drill cuttings collected in the 1980s. However, the two T_max values >460 °C in Well A occur within the same stratigraphic interval and coincide with consistent HI and TOC trends (Figure 7). Direct H₂ measurements are required to confirm free-H₂ release.

5.3 Hydrogen migration and trapping

The study area contains three generation mechanisms: radiolysis, water–rock interaction involving Fe(II)-bearing minerals in the granitic–gneissic and mafic basement, and thermogenic H₂ generation from sufficiently mature organic matter. Assessing the subsurface hydrogen system requires source-to-reservoir connectivity, sufficient reservoir capacity, effective trapping and sealing, and charge rates that exceed physical, geochemical, and microbial losses.

CNN-based interpretation delineates NNE–SSW striking basement faults that may connect basement H₂ sources with the overlying sedimentary succession. Volcanic structures increase local structural complexity and may further enhance vertical connectivity. The dip-azimuth distributions differ between the two structural domains. In the sedimentary basin, azimuths cluster between 11.25° and 45° (Figure 4D), indicating a preferred fault orientation that may channel fluid migration. In contrast, the volcanic structural high shows a broader, multimodal dip-azimuth distribution between 45° and 135° (Figure 4E), indicating multiple fault orientations and greater structural complexity. This pattern may increase fault intersections and vertical connectivity, suggesting that the volcanic rock could provide migration pathways where the faults remain hydraulically open.

H₂ migration is controlled by pore-throat size, capillary entry pressure, permeability, and water saturation. Advection is favored in permeable sandstone and open fractures, whereas diffusion becomes more important in fine-grained, low-permeability rocks such as shale and evaporite (Hutchinson et al., 2024; Lodhia et al., 2024). A two-stage charge model is therefore proposed for the study area, with vertical transport along connected basement faults followed by lateral redistribution through sandstone reservoir units. Basin-scale numerical models demonstrate the potential for long-distance lateral transport along connected aquifers, but do not constrain migration distances or rates in the Jeju Basin (Nicoletti et al., 2025). Accordingly, fluxes and travel times from analogue fault systems should be treated as order-of-magnitude estimates rather than direct measures of reservoir charge. Overlying shale and siltstone intervals may provide seal facies, with retention controlled by lateral continuity, thickness, capillary entry pressure, diffusivity, water saturation, and through-going faults or fractures.

As illustrated in the conceptual Model (Figure 9), the overlying siltstones and shales may restrict H₂ migration. Diffusive transport through fine-grained rocks is generally slower than advective migration through permeable faults and reservoirs (Hutchinson et al., 2024; Etiope et al., 2026). H₂ accumulation would therefore depend on the balance between reservoir charge and losses through diffusion, fault leakage, dissolution, and geochemical or microbial consumption. Quantitative assessment requires constraints on fault transmissibility, reservoir pore volume and water saturation, seal capillary entry pressure and diffusivity, H₂ solubility, and consumption rates. Within these limitations, the integration of seismic interpretation, fault attributes, basement lithology, and generation estimates provides a testable geological framework for natural hydrogen exploration in the northern East China Sea Shelf Basin.

6 Conclusion

This study integrates 2D and 3D seismic data, wireline logs, and geochemical analyses to evaluate natural hydrogen prospectivity in the Jeju Basin and Hupijiao Rise of the northern ECSSB. CNN based fault-detection delineates basement-involved NNE–SSW-striking faults and igneous intrusions that may provide vertical migration pathways from the crystalline basement to the overlying sedimentary sequences. Following vertical transfer along connected faults, H₂ may be redistributed laterally through sandstone reservoir units, whereas shale and siltstone intervals may restrict upward migration where they remain continuous and unfaulted.

Wireline logs identify fractured, water-saturated granitic and granodioritic intervals that provide conditions for radiolytic H₂ generation and water–rock interaction. Rock-Eval and elemental data from Well A record thermal overprinting near the upper limit of the mixed-generation stage (T_max >460 °C; R_o ≈1.1%), indicating mobilization of structurally bound hydrogen rather than free-H₂ expulsion. Time-equivalent strata buried below ∼2,800 m in adjacent depocenters may have reached the active-release stage and the post-mature window for free H₂ generation. A cumulative H₂ generation framework is proposed that integrates radiolytic and water–rock interaction contributions.

The integration of structural interpretation, wireline logs, and geochemical data provides a basis for evaluating natural hydrogen prospectivity in continental margin basins. Quantitative assessment will require improved constraints on thermal maturity, basement composition, reactive mafic-rock volume, fault transmissibility, reservoir properties, and H₂ preservation. Future work should focus on refining the quantitative parameters of the cumulative H₂ generation potential by incorporating basin-scale thermal-maturation modeling.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Huan Li, Central South University, China

Reviewed by: Muhammad Imran Rashid, University of Engineering and Technology, Lahore, Pakistan

Relly Margiono, Sekolah Tinggi Meteorologi Klimatologi dan Geofisika, Indonesia